Bicycle path planning in Johannesburg: aggregating user-defined spatial criteria to create efficient routes for bicycle infrastructure

Abstract

Recent initiatives by the city of Johannesburg to increase non-motorised transport through the installation of bicycle infrastructure were conducted without consulting the cycling preferences of the public. This study distributed a cycling preference survey, achieving fair reliability using the weighted Kappa statistic, in which potential users indicated the most important spatial factors for ideal cycling routes through Likert-scale answers. Importance rankings derived by Likert sums were combined with variability-explaining rankings derived by modified principal component analysis using polychoric correlation coefficients to produce a final list of retained spatial variables. These variables were quantified using secondary spatial data sets which were dichotomized into Boolean operators for network attributes in ArcGIS Network Analyst. The solved routes using the spatial factors derived by survey respondents were significantly different from the simple shortest-path routes between pre-defined origin and destination nodes. Shortcomings in the directness of the solved routes qualify their use as an initial step for non-motorised transport planning rather than a strict, unmodifiable route for bicycle lanes. Further experimentation with higher quality spatial data, custom routing algorithms, and a larger survey population may yield improved results in the future. The incorporation of local cyclists and future cyclists are a key factor in bicycle route design that should be included in non-motorised transport planning.